Efficient Machine Learning for Malnutrition Prediction among under-five children in India

Saksham Jain, Tayyibah Khanam, Ali Jafar Abedi, Abid Ali Khan

Abstract:

Child malnutrition is considered to be one of the
leading causes of infant mortality and malnutrition. This study
was aimed to leverage the advantages offered by machine learn-
ing models in terms of determining and accurately predicting
significant factors of malnutrition. For this study, the Children’s
recode files from the Indian Demographic and Health Survey
(IDHS) datasets from 2005-2006 and 2015-2016 were used. To
examine the nutritional status of children aged 0-59 months,
this study looks at stunting (Height-for-age), wasting (Weight-for-
Height), and concurrent stunted wasting (Height-for-age-Weight-
for-Height). Regular Machine Learning models, Tabular Deep
Learning frameworks, H2O base models, and AutoML models
are the four types of machine learning models employed in
our research. This research found that Automated machine
learning algorithms and Tabular Deep Learning frameworks,
in general, outperformed other models in terms of speed and
efficiency, as well as Accuracy (up to 96.46%) and AUC-ROC
scores (up to 99.95%), which are important in classification
problems like this one. Following a graphical representation of
the importance of numerous drivers of malnutrition for all three
anthropometric indices, we concluded our findings by comparing
the performances of several models and determining the top-
performing algorithms. This paper significantly contributes to
the possibilities of using machine learning in identifying probable
correlates of malnutrition for the effective prevention, cure, and
identification of target groups.